PSI - Issue 11

V. Gazzani et al. / Procedia Structural Integrity 11 (2018) 306–313 Gazzani et al. / Structural Integrity Procedia 00 (2018) 000–000

309

4

Each level was instrumented at least in two corners. At each corner, two high sensitive accelerometers, measuring in two orthogonal directions, were placed. Other couples of accelerometers were put in different positions at various levels to obtain more information about the dynamic behaviour of the whole tower. A wired sensor network was used, composed of two types of piezoelectric sensors (Integrated Electronic Piezoelectric - IEPE):  KS48C-MMF with voltage sensitivity of 1V/g and measurement range of ±6g;  KB12VD-MMF with voltage sensitivity of 10V/g and measurement range of ±0.6g. The digital recorder (DaTa500) is composed by a 24-bit Digital Signal Processor (DSP), an analogue anti aliasing filter and a high-frequency acquisition range (0.2Hz to 200kHz). RG58 coaxial cables link accelerometers and recorder. M28 and M32 signal conditioners with a frequency range of 0.1 to 100kHz and selectable gain are also used for each record. Some images of the used equipment for AVS are reported in Fig. 2. 3.2. Dynamic measurement and estimation of structural dynamic properties The collected measurements were initially sampled at 1000 Hz. A factor of 40 decimated them before processing to have the final data of 12.5 Sample per Second (SPS) (Singh et al., 2014). The record duration varied between forty minutes to one hour: it should be long enough to eliminate the influence of possible non-stochastic excitations during the test (McConnell, 2001). The described procedure was used for each AVS. A modal parameter extractor developed in Labview ® programming language carried out data processing. It can perform analyses in the time domain according to the SSI-Cov procedure mentioned before. The stabilization diagram obtained from the analysis of the collected data through Cov-SSI is reported in Fig. 3.

Fig. 3. Stabilisation diagrams (Cov-SSI) for Pomposa bell tower in Codigoro (Ferrara).

It shows the alignments of stable poles, for increasing order models, and it allows the determination of the n eigenvectors of dynamic matrix [ A ] which are representative of structural modes, and how many are instead purely numeric (due to their redundancy of calculation or noise). Red points indicate negative results of the stability test, while green points represent the positive ones: considering that natural modes show intrinsic characteristics of the towers, they are invariant to the process and the order of the NM (Van Overschee and De Moor, 1996b). Then, it is possible to isolate the natural modes from the numerical ones by increasing the order of the model and checking the stability of the results. The stability of a pole is defined as follow:  the estimated frequency is considered stable if it does not change more than 2%;  the damping for different orders should not deviate more than 15%;  the modal shape obtained by a certain order is compared to the same one obtained by a minor order by Modal Assurance Criterion (MAC) that must be at least equal to 90%. The identified frequencies appear slightly spaced and local modes show only for values upper to the 3 rd (Fig. 3).